playwright-best-practices

Flaky Playwright tests are usually caused by race conditions, explicit waits, or state leakage between tests. Replace waitForTimeout with web-first assertions and auto-waiting, ensure test isolation through fixtures, and use the trace viewer to diagnose timing issues.

Prefer user-facing locators like getByRole, getByLabel, and getByText over CSS or XPath selectors, since they reflect how users interact with the page. Avoid brittle selectors tied to implementation details that break on minor UI changes.

Yes, Playwright integrates with @axe-core/playwright to run automated accessibility scans. You can analyze full pages or scoped regions, filter by WCAG tags, and build a reusable fixture that fails tests when violations are detected.

Playwright supports both, with constraints. Browser extensions require Chromium launched with a persistent context in headed mode, while Electron apps are automated by connecting to the Electron main process and accessing renderer windows.

Failures limited to CI or parallel execution typically indicate shared state, resource contention, or timing differences under load. Isolate test data per worker, avoid global mutable state, and use proper waiting instead of fixed timeouts.

Use Page Object Model when multiple tests interact with the same pages or components, to centralize locators and actions. It reduces duplication and makes tests resilient to UI changes, but adds overhead for very small suites.